[Performance] The performance of dgl.DGLGraph.apply_edges has dropped in some test.

## 🐛 Bug

`apply_edges` on cpu device has performance drop recently.

The code for this api is in `python/dgl/heterograph.py`.

The result of time consumption is as follow:
api.bench_builtin_apply_edges 'cora' 'coo' 128 'u->e' 0.0005874281500013012 standard_result: 0.0002504343749990312
api.bench_builtin_apply_edges'cora' 'coo' 128 'u+v' 0.0005175077500013004 standard_result: 0.0002796029500018449
api.bench_builtin_apply_edges 'cora' 'coo' 512 'u->e' 0.0015405381999983093 standard_result: 0.0007349309500000345
api.bench_builtin_apply_edges'cora' 'csr' 128 'u+v' 0.0008072346749997905 standard_result: 0.0003226958249996414
api.bench_builtin_apply_edges'cora' 'csr' 512 'u->e' 0.0019674952999992 standard_result: 0.000725922425002068

## To Reproduce

Steps to reproduce the behavior:

1. we can get the result from regression test.

## Expected behavior

## Environment

- DGL Version (e.g., 1.0):
- Backend Library & Version (e.g., PyTorch 0.4.1, MXNet/Gluon 1.3):
- OS (e.g., Linux):
- How you installed DGL (`conda`, `pip`, source):
- Build command you used (if compiling from source):
- Python version:
- CUDA/cuDNN version (if applicable):
- GPU models and configuration (e.g. V100):
- Any other relevant information:

## Additional context